liveBook · Manning

Myth of average

As explained in Outlier Detection in Python

The myth of average is the idea that objects or observations can be typical in each individual dimension without any one being typical across all dimensions.

The myth of average says that an observation can be average on each individual dimension without any observation being average across all dimensions. Few multidimensional objects are completely typical. section 4.12

Fitting Cockpits to the Average Pilot

The US Air Force fitted cockpits to average height, arm length, and other measurements. Most pilots were average in most dimensions, but none was average in every dimension. section 4.12

In tabular data, enough features can make every row unusual from some angle. Outlier scores may therefore be non-zero for nearly every record, without a sharp boundary between inliers, noise, and outliers. section 4.12

How Outlier Detection in Python covers it

How deep it goes
Explains the intuition, dimensional consequence, and score interpretation but does not develop a formal detection method.
Examples
Uses cockpit fitting, mugs, and feature-rich tabular rows to illustrate multidimensional non-typicality.
What you can do after
Evaluate/understand why non-zero outlier scores may be legitimate for many records and why score ordering matters.
How it connects
Connects the myth of average to outlier detection, statistical outliers, noise, and the curse of dimensionality.
Cover of Outlier Detection in Python
liveBookmanning.com
Read Outlier Detection in Python on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Outlier Detection in Python. The section links point to the parts of the book it draws on.